Start with your ledger.
Track expenses and income in one place. Add your first entry yourself. No AI agent required.
Start your ledgerYOUR MONEY IN. YOUR MONEY OUT.
Your expenses.
Your income.
One clear ledger.
Add entries yourself, capture a receipt, or connect your AI agent. You decide what counts.
Built for freelancers and people working for themselves.
ONE REQUEST. YOUR FINAL SAY.
Proof EntryIllustrative demo · USD
01Capture02Review03Record
YOUR AI AGENTEXAMPLE REQUEST
“Log these expenses.”
Software 49.00Train ticket 28.50
AI AGENT · SUBSCRIPTION
Design software
49.00 USDCheck the details before recording.
Preparing proposalAI AGENT · TRAVEL
Train ticket
28.50 USDCheck the details before recording.
Preparing proposalConfirmed expensesNeither sample entry counted yet
1,284.00 USDYour AI agent captures the details.
Fictional example · No live connection · No automatic confirmationONE CONNECTION.
WHEN YOU WANT AN AGENT.
ClaudeChatGPTCursorVS CodeAny MCP client
01 / LESS ADMIN, MORE CONTEXT
A short path from
capture to record.
Capture your way.
Enter the details manually, upload a receipt, or ask your connected AI agent to do the entry work.
Manual entry · Receipt photo · AI agent
Keep the final say.
Review your agent’s proposals in Proof Entry. Only confirmed records count toward your totals.
Your review. Your record.
02 / BRING YOUR OWN AI AGENT
Use AI when you want to.
Manual entry works on its own. If you already use an AI assistant, connect it to capture entries for you to review.
Open connection setupOptional setup. No connection is needed to start your ledger.
Your expenses and income stay in one place. You decide what counts.
Optional: connect your AI assistant
MCP CONNECTIONHTTP
Use the setup instructions for your AI client, then complete the sign-in flow.
YOUR CLIENT, YOUR CONNECTION
Connect your AI agent.
Choose your client, follow its setup instructions, then approve the sign-in.
Terminal
claude mcp add --transport http proofentry https://staging.glidzr.com/api/mcp- 1Run it once. Add --scope user to have it in every project.
- 2Then /mcp inside Claude Code to sign in. A browser tab opens, you approve, and the client holds the token from then on.
Connecting a CLI without a browser callback? Use a device code.
Start your ledgerFree plan available. No card required.
Explore the tools, HTTP API and recorded exchanges
What that looks like, frame by frame
A recording, not a live connection — but the payloads are the real response shapes. Step through it; arrow keys work once it has focus.
You, in your client
What did I spend on travel last month?
1 / 4 · You ask
A plain question in your own client. No syntax, no app open.
What it can do — and what it never will
list_transactions- List ledger entries, newest first — expenses, income, or both.
get_totals- Totals for a period: money in, money out, and the net between them, plus the VAT position and what is still unbilled to clients.
totals_by_category- Confirmed totals grouped by category for a period, largest first.
list_categories- The category vocabulary for this ledger, split by which side it belongs to.
log_expenseproposes by default- Propose money SPENT.
log_incomeproposes by default- Propose money RECEIVED.
add_expenserecords by default- Add money SPENT.
add_incomerecords by default- Add money RECEIVED.
list_pending_review- Entries captured automatically that are waiting for the account holder to confirm or reject.
list_recurring- Standing instructions that write to this ledger on a schedule — rent, retainers, subscriptions.
assign_projectnever- A human chooses which engagement an entry belongs to.
create_recurring_rulenever- A standing instruction keeps writing on its own — more than the one reviewable row an agent is trusted with.
Logging defaults to proposals; direct add defaults to confirmed. Both accept confirm: true or false. Confirmed entries require complete structured data.
Or just HTTP
A cron job, a shell script, or an agent framework that has never heard of MCP can post an entry with the same token. It defaults to the review queue. Pass confirm: true with complete structured data to record it directly.
curl -X POST https://staging.glidzr.com/api/transactions \
-H "Authorization: Bearer $TOKEN" \
-H "Content-Type: application/json" \
-d '{"direction":"income","description":"Invoice 118","amount":"4500","currency":"AED"}'Amounts go in as strings and come back as integer minor units with the currency exponent, so nothing has to assume two decimal places. They are always positive; `direction` is what says which way the money went. Leave it out and you get an expense. The same facts, written for your agent, live at /llms.txt.
03 / BUILT AROUND YOUR WORK
For the person who does
the work. And the admin.
Money in. Money out.
Keep income and expenses in view, with clear direction and period totals.
The costs that repeat.
Keep track of recurring instructions for things like rent, retainers, and subscriptions.
More context per entry.
Keep currencies, categories, and client-rebillable costs connected to the record.
A FEW DETAILS
Before you start.
What do you mean by “AI agent”?
The AI assistant or client you connect to Proof Entry—for example, an MCP-capable client you already use. It can use Proof Entry’s tools according to the permissions it receives.
When do I need to review an expense?
Chat, receipt and email capture wait for review. Connected agents can propose entries or explicitly record complete structured data. Confidence alone never confirms an entry.
Do I need an AI agent?
No. You can enter expenses and income yourself and use your ledger without connecting an agent. Receipt photo, forwarded email, chat, and an optional AI connection give you other ways to capture.
Is this page connected to my account?
The demonstrations use fictional data in your browser and never write to your account. To connect your own AI agent, follow the setup instructions and approve its sign-in.
Your ledger. Your way.
You decide what counts.